eval'd expressions inside begin..end should be top-level when possible

[mlubin]

Looks like a bug in eval:

   _       _ _(_)_     |  A fresh approach to technical computing
  (_)     | (_) (_)    |  Documentation: http://docs.julialang.org
   _ _   _| |_  __ _   |  Type "help()" to list help topics
  | | | | | | |/ _` |  |
  | | |_| | | | (_| |  |  Version 0.2.0-514.r3d43
 _/ |\__'_|_|_|\__'_|  |  Commit 3d43df9032 2013-03-16 06:59:07
|__/                   |

julia> @eval begin
           type T 
               x 
           end
           T(sum([1,2]))
       end
T(3)

julia> @eval begin
           type T2 
               x 
           end
           s = 0
           for i in [1,2]
               s += i
           end
           T2(s)
       end
ERROR: error compiling anonymous: unsupported or misplaced expression composite_type in function anonymous

[JeffBezanson]

You can put let ... end around the for loop as a workaround.

[mlubin]

That works, thanks.

[mlubin]

It seems the issue is a bit more fundamental:

julia> function f2()
           type T
               x
           end
           for i in 1:2
               println(i)
           end
       end
# methods for generic function f2
f2() at none:2

julia> f2()
ERROR: error compiling f2: unsupported or misplaced expression composite_type in function f2

I changed the title. I couldn't find any open issues for this.

[mlubin]

@JeffBezanson, can you clarify if this is a bug or a "feature"? I'll close the issue if it's the latter.

[JeffBezanson]

It's neither a bug nor a feature; it's just that there are no "local types". Types have to be defined at the top level. In the original issue, there are no explicit scopes introduced around the type definition, so it's reasonable to expect that it's at the top level.

[mlubin]

Got it. The error message for attempting to define types inside functions could also be improved.